Member of Technical Staff - LLM Pre-Training
Core
Design and optimize large-scale LLM pre-training systems, including data pipelines, model architectures, and training algorithms for on-device deployment.
Role type
Staff / Senior IC LLM Pre-Training Researcher and Engineer
Builds
Pre-trained LLMs optimized for on-device inference and hardware efficiency
Domain
Artificial Intelligence, Large Language Models, Deep Learning, On-device AI
Deliverable
production ML models
Required skills
LLM architecture design, deep learning optimization, distributed training (FSDP, TP, PP, CP), low-level system optimization, Python, PyTorch
Preferred skills
PhD in relevant field, publication record in top venues (NeurIPS, ICML, ICLR), experience with 100k+ GPU hour budgets
Technologies
PyTorch, FSDP, TP, PP, CP, GPU clusters
Responsibilities
Develop and evaluate novel ideas across architecture, optimizers, and training algorithms; make training run efficiently on multi-node GPU clusters; co-design architectures with hardware optimization teams; publish at top venues
Seniority
Staff / Senior, hands-on IC